Christoph Höger wrote: > Hi, > > anyone knows about a (high-performance) regular expression to match > java-like Strings? > (e.g. "Hi, World \n this is a \"-quoted string.\n") > > I have tested > > ((\\.)|[^"\\])* > > which basically does what I want (although capturing too much escape > sequences). > > The problem is: I've tried jakarta's regexp and java's implementation > and both run into Stack Overflows for input strings with more than 500 > characters. That is definitely not acceptable as a hard limit for > tokenizing source code. > > Any suggestions? > IIRC from taking a compiler course, tokenizing is usually done with a lexer like Lex, Flex, JLex, Ragel, and so on (in conjunction with Yacc or Bison). They're highly tuned for the task - for example, Ragel creates a state machine (decision trees), and only needs to scan each character of the source text once. This is as high performance as you can expect to get (maybe you can squeeze a cycle or two more by coding in assembly, of course). See http://www.complang.org/ragel/ I'm not sure exactly how regular expressions work, but I suspect they're not nearly this robust. HTH, Konstantin -- fedora-list mailing list fedora-list@xxxxxxxxxx To unsubscribe: https://www.redhat.com/mailman/listinfo/fedora-list Guidelines: http://fedoraproject.org/wiki/Communicate/MailingListGuidelines
